Relevance of breast cancer cell lines as models for breast tumours: an update
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Breast Cancer Research and Treatment 83: 249–289, 2004.
© 2004 Kluwer Academic Publishers. Printed in the Netherlands.
Review
Relevance of breast cancer cell lines as models
for breast tumours: an update
Marc Lacroix and Guy Leclercq
Laboratoire Jean-Claude Heuson de Cancérologie Mammaire, Institut Jules Bordet, Université Libre de Bruxelles,
Bruxelles, Belgium
Key words: breast cancer, cell lines, classification, estrogen receptor, gene expression, Her-2/neu, markers, models,
tumours
Summary
The number of available breast cancer cell (BCC) lines is small, and only a very few of them have been extensively
studied. Whether they are representative of the tumours from which they originated remains a matter of debate.
Whether their diversity mirrors the well-known inter-tumoural heterogeneity is another essential question. While
numerous similarities have long been found between cell lines and tumours, recent technical advances, including
the use of micro-arrays and comparative genetic analysis, have brought new data to the discussion. This paper
presents most of the BCC lines that have been described in some detail to date. It evaluates the accuracy of the
few of them widely used (MCF-7, T-47D, BT-474, SK-BR-3, MDA-MB-231, Hs578T) as tumour models. It is
concluded that BCC lines are likely to reflect, to a large extent, the features of cancer cells in vivo. The importance
of oestrogen receptor-alpha (gene ESR1) and Her-2/neu (ERBB2) as classifiers for cell lines and tumours is under-
lined. The recourse to a larger set of cell lines is suggested since the exact origin of some of the widely used lines
remains ambiguous. Investigations on additional specific lines are expected to improve our knowledge of BCC and
of the dialogue that these maintain with their surrounding normal cells in vivo.
Introduction surrounding stroma. Most of the available BCC lines
issued from metastatic tumours, mainly from pleural
A considerable part of our knowledge on breast effusions. Effusions provided generally large numbers
carcinomas is based on in vivo and in vitro studies of dissociated, viable tumour cells with little or no
performed with breast cancer cell (BCC) lines. These contamination by fibroblasts and other tumour stroma
provide an unlimited source of homogenous self- cells. However, even with metastatic samples, suc-
replicating material, free of contaminating stromal cess in long-term propagation has been limited. For
cells, and often easily cultured in simple standard me- instance, Cailleau et al. [4], Meltzer et al. [5], and
dia. The first line described, BT-20, was established Gazdar et al. [2] fruitfully propagated tumour cells in
in 1958 [1]. Since then, and despite sustained work in only 10, 2, and 25% of cases, respectively.
this area, the number of permanent lines obtained has Many of the currently used BCC lines were estab-
been strikingly low (about 100). Indeed, attempts to lished in the late 1970s. A very few of them, namely
culture BCC from primary tumours have been largely MCF-7, T-47D, and MDA-MB-231, account for more
unsuccessful. For instance, Gazdar et al. [2] ob- than two-thirds of all abstracts reporting studies on
tained cell lines from only 18 of 177 primaries, while mentioned BCC lines, as concluded from a Medline
the percentage of success reported by Amadori et al. (http://www.ncbi.nlm.nih.gov/PubMed/)-based survey.
was as low as 0.7 (1/136) [3]. This poor efficiency The transposability to tumours of results obtained with
was often due to technical difficulties associated such limited numbers of cell lines is questionable. To
with the extraction of viable tumour cells from their discuss the problem of representativeness, we have250 M Lacroix and G Leclercq
Table 1. A series of BCC lines for which a somewhat more detailed description has been given in the literature
Cell line Type of Original Modal chromosome ER PgR Reference
cancer tissue number(s) status status
‘21-series’ [6]
AU565 IDC M (Pl) ? − − [7]
BOT-2 IDC P 63a ? ? [8]
BRC-230 IDC P 60–61a − − [3]
BrCa-MZ-01 MC P 66–70a + + [9]
BrCa-MZ-02 IC M (Pl) 46–50a − − [9]
BSMZ IDC M (Pl) 80a + + [10]
BT-20 IDC P 49b −g − [1, 11]
BT-474 IDC P 55a , 104c , 103d + + [11, 12]
BT-483 IDC P 72a + + [11, 12]
BT-549 PIDC P 74–76d , 74e − − h
CAL-18A C P 71a − − [13]
CAL-18B C P 65a − − [13]
CAL-51 IDC M (Pl) 46c,f − ? [14]
CAMA-1 C M (Pl) 78d , 80f + + [11, 15]
DU4475 IDC M (Sk) 87–90a , 88–93c , 91d , 93e − − [16]
EFM-19 IDC M (Pl) 62d + + [17, 18]
EP IDC M (Pl) 53a + ? [19]
EVSA-T IDCS M (As) 84b , 62c − + [11, 20, 21]
GI-101 IDC R (L) 98–100a − − [22]
GCS IDC M (As) ? + + [23]
HBL-100 See text 63a − − [24, 25]
‘HCC-series’ [2]
HDQ-P1 IDC P 55–59a , 92–107c − − [26]
HH315 C M (O) 113a − − [27]
HH375 C M (LN) 64 and 67a − − [27]
‘HMT-series’ See text [28]
Hs578T CS P 58a , 59e − − [11, 29]
Ia-270 IDC M (Pl) ? + + [30]
IBEP-1 IDC M (Pl) 52a − + [31]
IBEP-2 IDC M (Pl) 74a + − [31]
IBEP-3 IDC M (Pl) 57a − + [31]
IIB-BR-G IDC P 56a − − [32]
JCK IDC M (Pl) ? + + [23]
KPL-1 IDC M (Pl) 77–78a , 77c + − [33]
KPL-3C IDC M (Pl) 64a − − [34]
KPL-4 IDC M (Pl) 53a − − [35]
LCC15-MB C M (F) ? − − [36]
MA11 ILC M (BM) 64a − − [37, 38]
MAST IDC M (As) 60a + + [39]
MaTu IDC M (LN) 66–69b − − [40]
MCF-7 IDC M (Pl) 88a , 86c , 79d , 65f + + [11, 41]
MDA-MB-134 VI IDC M (Pl) 43a , 42d , 44 and 66f + − [4, 11, 42]
MDA-MB-157 MC M (Pl) 65a , 64–66b , 54 and 95d , 62, 116f − − [4, 43]
MDA-MB-175 VII IDC M (Pl) 49a , 48f + − [4, 11, 42]
MDA-MB-231 IDC M (Pl) 64a , 69–70b − − [4, 11, 42]
MDA-MB-330 ILC M (Pl) 64a − − [4, 11]
MDA-MB-361 AC M (Br) 52a , 54–56d , 51f + + [4, 11]Comparison of cell lines and tumours 251
Table 1. (continued)
Cell line Type of Original Modal chromosome ER PgR Reference
cancer tissue number(s) status status
MDA-MB-415 AC M (Pl) 72a ? ? [4, 11]
MDA-MB-435S IDC M (Pl) 64a , 56e , 57f − − [4]
MDA-MB-436 AC M (Pl) 45 and 80a , 39 and 80d − − [4, 11]
MDA-MB-453 AC M (Pl) 45a , 89c − − [4]
MDA-MB-468 AC M (Pl) 35a − − [4]
MFM-223 IDC M (Pl) 46–47a , 46d − − [44]
MPE-600 C ? 46d ? ? i
MT-1 C P (Tr) 103f − − [45]
MT-3 C P (Tr) 47c , 46f − − [45]
MW IDC M (Pl) 67a − ? [19]
PMC42 C M (Pl) 66a , 57f ? ? [46]
SK-BR-3 IDC M (Pl) 84a , 79d,f − − [11]
‘SUM-series’ See text [47]
T-47D IDC M (Pl) 66a , 63d , 65e , 62f + + [11, 48]
UACC-812 IDC P 58–64a , 63d − − [5]
UACC-893 IDC P 62a , 59d − − [5]
UISO-BCA-1 IDC M (Pl) 54a − − [49]
UISO-BCA-2 IDC M (Pl) 61a − − [49]
VHB-1 IDC P 70–74a + + [50]
ZR-75-1 IDC M (As) 71–72a , 71d , 72f + + [11, 51]
ZR-75-30 IDC M (As) 81a , 79f + − [11, 51]
a Original reference.
b Reference [11].
c Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) data.
d Reference [52].
e American Type Culture Collection (ATCC) data.
f Reference [53].
g ESR1 mRNA present.
h Coutinho W.G. and Lasfargues E.Y., 1978 (unpublished data).
i Developed by Vysis International Inc.
IDC: invasive ductal carcinoma; MC: medullary carcinoma; IC: inflammatory carcinoma; PIDC: papillary invasive ductal
carcinoma; C: carcinoma; IDCS = IDC, mucin-producing, signet-ring type; CS: carcinosarcoma; ILC: invasive lobular
carcinoma; AC: adenocarcinoma.
P: primary; M: metastasis; R: recurrence Pl: pleural effusion; Sk: skin; As: ascites; L: local; O: omenum; LN: lymph node;
F: femur; BM: bone marrow; Br: brain; Tr: transplanted.
here brought together and compared various data ob- or metastasis); the steroid receptor status and the
tained, mostly in the last decade, on both tumours and modal chromosome number(s) of the cell lines.
BCC lines.
Distinctive features of BCC lines. An exhaustive de-
Presentation of BCC lines – the question scription of all BCC lines contained in Table 1 is
of representativeness beyond the scope of the present paper. However, it
is of interest to mention that some aspect regarding
Multiplicity and variability of BCC lines their biology or their origin has distinguished many of
Most of the BCC lines that have in the past been the them. For instance, DU4475 cells may grow in sus-
subject of a somewhat detailed description are listed pension in vitro, a feature rarely observed with BCC
in Table 1. Also provided, where available, are data on [16]. KPL-3C cells may produce tumours associated
the type of primary tumour (ductal, lobular, . . .); the with micro-calcifications in nude mice [34]. CAL51
tissue from which the BCC lines originated (primary cells exhibit a normal karyotype [14] and appear252 M Lacroix and G Leclercq
perfectly diploid by molecular cytogenetic analysis metastasis [36] and HH315 and HH375 from abdom-
[53]. IBEP lines differ from the widely used MCF-7 inal and supraclavicular lymph node metastases, re-
and MDA-MB-231 cells by their spectrum of pro- spectively [27]. MA11 have been obtained from a bone
teolytic activities [54]; they are also characterised by marrow aspirate [37]. MAST cells have been isolated
a relatively rare steroid receptor status (two of them from an ascitic fluid [39], as were ZR-75-1 cells [51].
are estrogen receptor-negative (ER−)/progesterone HBL-100 cells were established from an early lacta-
receptor-positive (PgR+), the third is ER+/PgR−) tion sample obtained in an apparently healthy woman.
[31]. MFM-223 cells have a large amount of androgen They exhibited characteristics of transformation from
receptors [44]. The epidermal growth factor receptor the very beginning and evolved during in vitro main-
gene (EGFR) is amplified in BT-20 and MDA-MB- tenance, until becoming tumourigenic in nude mice.
468 lines [55]; it is over-expressed without ampli- They have been shown to harbour SV40 genetic in-
fication in SUM-102, SUM-149, and SUM-229 cells formation [24, 25]. HBL-100 cells have also been
[56, 57]. The fibroblast growth factor receptor 1 gene shown to carry the Y chromosome (ATCC web site),
(FGFR1, at 11q13) is amplified in SUM-44 and SUM- which raises some doubt as to its origin.
52 cells [58], and highly amplified in MDA-MB-134 Finally, some BCC lines are characterised by their
cells [59]. The fibroblast growth factor receptor 2 tropism to specific metastatic sites. This is the case for
gene (FGFR2, at 10q26) is amplified in SUM-52 MT-1 cells, which tend to specifically give bone/bone
[58, 60]. BT-474, SK-BR-3, MDA-MB-361, MDA- marrow metastases in mice, while MA-11 preferen-
MB-453, ZR-75-30, UACC-812, UACC-893, BSMZ, tially establish in the brain, following injection into
HCC1419, HCC1954, SUM-190, and SUM-225 lines the left-ventricle of the heart [65].
have an amplified ERBB2 (encoding Her-2/neu) at
17q11.2-q12 [2, 10, 58, 61, 62]. Close to this re-
BCC lines series. Some BCC lines are related, as
gion, UACC-812 cells have an amplified TOP2A gene,
they have been obtained from the same patient, or
while one copy of this gene is deleted in the MDA-
isolated in the same laboratory (and thus often estab-
MB-361 line; as a consequence, the sensitivity of both
lished and maintained in the same culture conditions).
cell lines to topoisomerase II alpha inhibitors is mod-
Other lines have been derived by serial subculture
ified [61]. One of the most intriguing cell lines is
from the same initial population. These cell lines
PMC42 [46], which is apparently derived from the
constitute series that are well suited for comparative
stem cell compartment in the breast. These cells are
studies. ‘HCC’, ‘SUM’, ‘HMT’, and ‘21’ cell lines are
heterogeneous, with at least eight different morpholo-
examples of such series.
gical types identified by phase contrast and electron
microscopy, expressing both secretory and myoep- ‘HCC (Hamon Cancer Centre) series’. In an at-
ithelial markers. These cells can express milk-specific tempt to obtain paired tumour and non-tumour cell
genes through hormone and extra-cellular matrix in- lines from patients with breast cancer, 18 BCC lines
teractions [63]. Moreover, they can be induced to were isolated from 177 primary tumours (and 3 from
undergo epithelial–mesenchymal transition (EMT, see 12 metastatic lesions) [2]. This series allowed an
below) [64]. Both of these attributes are quite unusual extensive comparison of various features (morpho-
among the cell lines. logy, ploidy, and marker expression) in paired BCC
AU565 and SK-BR-3 lines were obtained from lines/original tissue [66]. The high number of ori-
the same patient [7], as were CAL18A and CAL18B ginal tissue samples also allowed the identification
[13]. MDA-MB-330 and MA-11 issued from a lobu- of criteria leading to the most successful isolation of
lar carcinoma, a type representing only 5–10% of all BCC lines, thus pointing out that cell line establish-
breast carcinomas [4, 11, 37, 38]. Hs578T cells were ment is a biased process (see below).
derived from a carcinosarcoma, a very rare form of ‘SUM series’. A series of 11 tumour cell lines that
breast cancer [29], as is medullary carcinoma, from were isolated and grown in media improved for the
which BrCa-MZ-01 and MDA-MB-157 originated culture of normal breast epithelial cells. They were
[9, 43]. obtained from primary tumours, pleural effusions,
Besides the number of cell lines that have been a chest wall recurrence, or from a highly invasive
obtained from pleural effusions, others have a more cancer specimen grown for two transplant gener-
infrequent origin: MDA-MB-361 issued from a brain ations in immuno-deficient mice. Molecular cyto-
metastasis [4, 11], LCC15-MB cells from a femoral genetic analysis of these BCC lines was performedComparison of cell lines and tumours 253
[47, 58]. A detailed description of phenotypic/ from MDA-MB-435S [74]. Distinct sub-populations
genotypic features of each SUM line is access- of MDA-MB-231 BCC have been obtained on the
ible (http://www.cancer.med.umich.edu/breast_cell/ basis of their preferential metastatic site (bone or
clines/clines.html). brain) [75].
‘HMT series’. The HMT-3522 cell line was de-
rived from a fibrocystic lesion of the breast. Issuing BCC lines from breast cancer patients with germ-
from breast tissue cultured on type IV collagen, line mutations. A particular subset of tumours is
these cells were first propagated as a near-diploid, composed of those arising in women with germ-line
non-tumourigenic sub-line in enriched but serum- mutations. Attempts to isolate corresponding cell lines
free medium. These cells progressively developed, appear to have been mostly unsuccessful.
by subcultures in various conditions: p53 mutation; To the best of our knowledge, only one BRCA1
MYC amplification; EGF-independence accompa- mutant cell line, HCC1937, has been described to
nied by tumourigenicity in nude mice; EGFR, TGF- date [76]. It was established from a primary non-
alpha, and c-erbB-2 over-expression. They have also metastatic IDC originating from a 24-year-old patient
been genetically engineered to study inter-relations with a germ-line mutation. The cell line is homo-
between EGF and 17-beta oestradiol (E2 ) action (for zygous for the BRCA1 5382insC mutation, whereas
a review, see [28]). the patient’s lymphocyte DNA was heterozygous for
‘21-series’. Two primary and two metastatic cell the same mutation, as were at least two other fam-
lines with distinct phenotypes and genotypes estab- ily members’ lymphocyte DNA. HCC1937 BCC also
lished from the same patient diagnosed as having have an acquired mutation of TP53 with wild-allele
infiltrating and intra-ductal mammary carcinoma [6]. loss, and an acquired homozygous deletion of the
PTEN gene. No significant levels of progesterone or
BCC variant lines. A number of variant sub-lines oestrogen binding were observed in either the primary
have been obtained, starting from the most widely tumour or the HCC1937 cultured cells. Only very low
used BCC lines. Some resulted from culture of the levels of Her-2/neu were expressed. HCC1937 cells
same unstable cell line by different groups and/or in have been extensively used to study BRCA1 function,
different culture mediums. For instance, it is well- notably after ionising radiation-induced damage [77–
known that MCF-7 cell stocks from different labora- 86]. Considering the number of distinct mutations that
tories may differ in their sensitivity to E2 , evaluated may affect the BRCA1 gene [87], there is clearly a
through cell proliferation rate and TFF1 (encoding need for additional cell lines. To be complete about
pS2) and CCND1 (cyclin D1) gene induction. These BRCA1, it must be mentioned that the establishment of
effects have been associated with variations in ER an immortalised breast cell strain containing the het-
protein and mRNA levels [67]. erozygous form of a BRCA1 185delAG mutation has
Other variants were obtained by selection of cell been described. These mutant cells appear to abund-
sub-populations resistant to a given agent, for instance antly express the 220-kDa full-size BRCA1 protein and
anti-estrogen, vitamin D, doxorubicin, thymidylate to have growth and stress response characteristics sim-
synthase inhibitor . . . [68–71]. A good example of ilar to those of normal human breast cells, which is
variant use to explore a specific resistance is provided consistent with the hypothesis that loss of heterozy-
by the MCF-7 BCC, which have been often considered gosity must occur to impair putative BRCA1 function
as ‘prototypes’ for ER+ cells. Different levels of [88].
MCF-7 resistance to (anti)-estrogens have been found, It seems that there is no available BRCA2 mutant
illustrated by different variants. So are MCF-7/LCC1 BCC line at the present time, suggesting that estab-
hormone-independent but hormone-responsive, MCF- lishment of such line from BRCA2 tumours could be
7/LCC2 (selected from LCC1) 4-hydroxytamoxifen especially difficult.
(4-OH-TAM, a partial anti-oestrogen)-resistant but ICI One cell line, HCC1569 [2], was found to have a
182,780 (a pure anti-oestrogen)-sensitive, and MCF- mutated FHIT gene (G → T at nucleotide 651, chan-
7/LCC9 (selected from LCC1) 4-OH-TAM- and ICI ging valine to phenylalanine) that proved to be her-
182,780-resistant [72, 73]. itable, in that the patient’s daughter also carried the
Specific in vivo properties have been associated to same alteration. The tumour arose in an older patient
a third category of variants. For instance, MDA-MB- (age 70) without a family history of breast cancer
435/LCC6 cells constitute an ascites model derived [89]. It is unclear whether the germ-line alteration that254 M Lacroix and G Leclercq
occurred in this breast tumour (which otherwise ap- expressed in IBC. RhoC is involved in cytoskeletal
peared to be sporadic) was a causative factor in the re-organisation; specifically, it is involved in the form-
development of her cancer. ation of actin stress fibers and focal adhesion contacts.
Li-Fraumeni syndrome (LFS) is a rare, familial, Its over-expression modulates induction of angiogenic
dominantly inherited cancer syndrome characterised factors in BCC. Treatment of cells with a farnesyl
by a wide spectrum of neoplasms occurring in chil- transferase inhibitor may lead to the reversion of RhoC
dren and young adults. While LFS has been associated GTPase-induced inflammatory phenotype. [93, 96–
to germ-line mutations in TP53, no cancer cell line 101]. Contrasting with ARHC, the expression of LIBC
has been as yet isolated from LFS-associated breast was found to be frequently lost in IBC. LIBC is now
tumours. However, the spontaneous in vitro immortal- renamed WISP3 (Wnt-1 inducible signalling pathway
isation of normal breast epithelial cells obtained from protein 3). It was shown that its loss of expression
a patient with LFS (with a mutation at codon 133 of may contribute to the phenotype of IBC by regulating
TP53) has been described [90]. tumour cell growth, invasion and angiogenesis [96,
Other germ-line mutations that have been suggest- 102]. WISP3 is a member of a gene family (‘CCN
ed to be associated with breast cancer susceptibility family’), also including CTGF (connective tissue
may be found in the PTEN, ATM, and NBS1 genes growth factor), CYR61 (cysteine-rich angiogenic in-
[91]. To our knowledge, no BCC line has been ob- ducer 61), and NOV (nephroblastoma over-expressed
tained from carriers of these types of mutations. gene), which encode cysteine-rich secreted proteins
with roles in cell growth and differentiation.
The specificity of inflammatory breast cancer. In- Although this article is mainly focused on cell
flammatory breast cancer (IBC) is an advanced and lines, we need to mention the existence of two xeno-
accelerated form of breast cancer usually not detected graft models of IBC. The first, MARY-X, grows as
by techniques such as mammograms or ultrasounds. It tight multi-cellular spheroids in vitro and as lympho-
requires immediate aggressive treatment with chemo- vascular emboli in vivo in SCID/nude mice (animal
therapy prior to surgery and is treated differently models for tumourigenicity studies are discussed in
than more common types of cancer. While no more the section Tumourigenicity of BCC lines in animal
than 5% of all patients with breast cancer have IBC models of this paper). It is ER−, PgR−, Her-2/neu−
in Western countries, this percentage may be higher and EGFR+. The primary tumour of origin of MARY-
than 10% in African countries such as Tunisia or X exhibited identical markers, except that about 50%
Nigeria [92]. The majority of IBC tumours are ER- of its cells showed Her-2/neu amplification. Com-
negative (ER−), PgR-negative (PgR−), Her-2/neu- parative studies of MARY-X with non-inflammatory
positive (Her-2/neu+), and EGFR-positive (EGFR+) xenografts indicated 10–20-fold over-expression of
[93, 94]. They are also characterised by intense an- E-cadherin and MUC1, findings that were reflect-
giogenesis and a strong E-cadherin expression [95]. In ed in most cases of human IBC. The formation of
view of their ER, PgR, and EGFR status, the presence spheroids and the lack of binding of the tumour em-
of a high E-cadherin level in IBC is intriguing and con- boli to the surrounding endothelium have been as-
trasts with observations in most other breast cancers sociated to: (a) an over-expressed E-cadherin/alpha,
(see notably the sections Phenotype and invasiveness- beta-catenin axis, determining strong homotypic
based BCC line classification and Analysis of breast cell interactions; (b) a decreased alpha-3 4-fucosyl-
tumours – markers and grade – comparison with cell transferase activity, which leads to reduced sialyl-
lines in this paper). IBC deserves a specific discus- Lewis X/A (sLe(X/A)) carbohydrate ligand-binding
sion here, as basic research on this form of cancer epitopes on the over-expressed MUC1 and other
has recently greatly benefited from the introduction of surface molecules that bind endothelial E-selectin.
human cell line and xenograft models. Moreover, the decreased sLe(X/A) fail to confer elec-
The SUM-149 (ER−, PgR−, Her-2/neu−, and trostatic repulsions between tumour cells, which fur-
EGFR+) and SUM-190 (ER−, PgR−, Her-2/neu+ ther contributes to the compactness of the MARY-X
(ERBB2 amplified), and EGFR+) cell lines are de- spheroids by allowing the E-cadherin homodimeric
rived from primary IBC tumours. The SUM-149 BCC interactions to go unopposed [103–105].
have been used to identify genetic determinants of WIBC-9 is another human xenograft transplant-
the IBC phenotype. Among the genes found, ARHC able in SCID/nude mice. It is frequently accompanied
(coding for the small GTPase RhoC) appears over- by lung metastasis and exhibited erythema of theComparison of cell lines and tumours 255
overlying skin, reflecting its human counterpart. In in different laboratories and/or cultured under vari-
vitro, WIBC-9 forms tube-like structures and loops, ous conditions could differently evolve and give rise
in concordance with its in vivo feature. Consistent to distinct sub-populations. This would prevent inter-
with IBC, WIBC-9 is ER−, PgR−, and exhibits laboratory comparisons of data obtained with the same
ERBB2 gene amplification. Comparative studies of line.
WIBC-9 and three established non-IBC xenografts, Finally, that the few widely used BCC lines could
by reverse transcription-PCR, ELISA, and immuno- accurately reflect the inter-tumoural heterogeneity has
histochemistry, indicated the over-expression of a been debated.
series of angiogenesis-related genes in IBC [106]. A response to these questions and criticisms will
be given throughout the present paper.
The problem of representativeness in BCC lines
Permanent BCC lines have been isolated in order to Phenotype and invasiveness-based studies
understand the mechanisms underlying tumour ini- of BCC lines and tumours
tiation and evolution. Yet despite the considerable
role that they continue to play in most aspects of Phenotype and invasiveness-based BCC line
cancer biology, they are still often viewed as non- classification
representative models of the tumours from which they As exemplified by the steroid receptor status, dis-
are derived. cordances have been pointed out between tumours
At first, the metastatic origin of most BCC lines and the panel of available BCC lines. Various ex-
has raised questions as to their relationship to primary planations have been proposed. One of them, based
tumours. More generally, the pure and clonal pop- on the epithelial–mesenchymal transition hypothesis,
ulation composing any single BCC line, regardless was suggested by initial observations of distinct cell
of its primary or metastatic origin, is a priori ex- line phenotypes. Subsequent studies confirmed the
pected to poorly reflect the assumed heterogeneity of existence of two major ‘portraits’ for BCC lines.
breast tumours. This heterogeneity is in line with the
common view that breast cancer involves a sequen- Steroid receptor status and the bias in BCC line
tial progression through clinical and pathologic stages, isolation. In the classification and comparison of cell
starting with atypical hyperproliferation, progressing lines (and breast tumours), no single criterion appears
into in situ, then invasive carcinomas and culminating a priori more pertinent than the ER. As a mediator
in metastatic disease. These changes are believed to be of (anti)-estrogen action, it plays a central role in
associated with the sequential acquisition of various breast cancer biology and treatment. One of the ma-
genetic and phenotypic changes in a single cell fol- jor proteins induced by estrogens is the progesterone
lowed by clonal selection and expansion, thus leading receptor (PgR). ER and PgR amounts have been eval-
to intra-tumoural diversity. uated in tumours and cell lines for more than 30 years
It has also been argued that most culture conditions [107]. ER was long believed to be unique; however,
relevant to the establishment of BCC lines will elim- an isoform named ER-beta and encoded by a specific
inate some types of tumour cells initially present in gene, ESR2, was identified in the late nineties. The
the cancer samples. For instance, BCC unable to grow ‘older’ ER isoform (renamed ‘alpha’, and encoded
well on plastic, or having an absolute requirement for by the ESR1 gene) seems to be functionally the most
factors only provided by their specific tumour envi- important in breast tumours, as no clear picture has
ronment, are unlikely to be represented in the panel emerged to date about the ER-beta role in this pathol-
of currently available BCC lines. Also, all else being ogy [108]. We have shown that the level of ER protein
equal, only the most proliferative cells extracted from evaluated in breast tumours by ligand-binding assay,
the tumour samples to be cultured in vitro should be which measures both ER-alpha and -beta isoforms,
finally selected. was linearly correlated to the level of mRNA specific
Cancer cells are genetically unstable. An addi- for ESR1, while the ESR2 mRNA was undetectable
tional criticism addressed to BCC lines is that, as in the samples [109]. In the present paper, we will
compared to cells in vivo, they could undergo specific consider that ER-beta, although not negligible (see
genotype/phenotype alterations resulting from long- notably [110]), is of secondary importance in breast
term culture in simplified conditions. In addition to tumours and cell lines, and the term ER will refer to
this divergence, lots of the same BCC lines distributed the alpha isoform, unless otherwise indicated.256 M Lacroix and G Leclercq
Cumulative data from a number of studies have this could explain the discrepancies between these
revealed that steroid receptors are distributed in lines and the primary tumours. The concept of phe-
breast tumours as follows: 50–60% ER+/PgR+; notype change in BCC, which is opposed to the idea
10–20% ER+/PgR−; 5–15% ER−/PgR+; 15–25% that the ‘portrait’ of tumour cells remains essentially
ER−/PgR−. In contrast, BCC lines listed in Table 1 the same during cancer progression, received a more
(and added with the HCC and SUM series) are char- precise formulation with the ‘epithelial–mesenchymal
acterised by a clearly different distribution: 20% are transition’ (EMT) hypothesis.
ER+/PgR+; 7% ER+/PgR−; 5% ER−/PgR+; 68% The EMT hypothesis was mainly based on studies
ER−/PgR−. One explanation for these discrepancies involving a relatively high number (up to 18) of BCC
supposes that the phenotype of BCC could change un- lines. These were found to distribute along a spec-
der in vitro culture, notably leading to the loss of ster- trum of differentiation from epithelial to mesenchymal
oid receptor expression. However, data presented in appearances [112, 113]. Based on their phenotype
various parts of this paper suggest that this is unlikely. and invasiveness (chemo-invasion through the recon-
On the other hand, it appears that steroid receptor- stituted basement membrane, Matrigel, in a modified
negative BCC lines are easier to establish in vitro Boyden chamber), the cell lines could, however, be
than the receptor-positive ones. The 18 cell lines of summarily classified into three groups:
the HCC series that were obtained from 177 primary
tumours [2] exemplify this. Only a subset of carcino- — The first group expressed high amounts of mark-
mas that had several features indicative of tumours ers typical of the luminal epithelial phenotype
with poor prognosis, absence of steroid receptors, hy- of breast cells: ER, E-cadherin (gene CDH1),
perploidy or aneuploidy, Her-2/neu over-expression, zonula occludens-1 (TJP1), and desmoplakin I/II
positive immuno-staining detection of p53 protein (DSP), the three latter being involved in adher-
expression, could be successfully cultured. Among ens, tight, and desmosomal junctions, respectively.
the 18 HCC lines, 15 (83%) are ER−. It has been These ‘luminal epithelial-like’ cells grew as inter-
suggested that the secretion of various extra-cellular connected colonies of polygonal cells on plastic
proteins, such as collagens, could provide a selec- and as fused colonies in Matrigel. They were
tive advantage to ER−/PgR− cells, by increasing their weakly invasive. BCC lines in this group included
adherence to plastic. It also appears that ER−/PgR− BT-483, MCF-7, T-47D, and ZR-75.
cells may, more frequently than ER+ and/or PgR+ — The second group of cell lines, closely related to
BCC, express both a series of growth factors (i.e., the first, was characterised by a ‘weakly luminal
EGF, TGF-alpha, amphiregulin, heregulin, FGFs, epithelial-like’ phenotype, with the expression, to
IGFs) and their corresponding receptors, thus sustain- a reduced extent, of at least some of the epithelioid
ing growth independently of exogenous growth factor markers found in the first group, and a weak in
supply (autocrine loop) [111]. vitro invasiveness. Most of these cell lines grew
as non-fused spheres in Matrigel. On plastic, they
Distinct phenotypes – the ‘epithelial–mesenchymal accumulated in clusters of loosely attached cells,
transition’ (EMT) hypothesis. To account for the reaching full confluency only rarely (personal ob-
under-representation of steroid receptor positivity in servations). In this group were included the BT-
BCC lines, another explanation has been advanced. 474, CAMA-1, MDA-MB-134, MDA-MB-361,
During the sequential in vivo progression of cancer MDA-MB-453, MDA-MB-468 and SK-BR-3 cell
from atypical hyperproliferation to metastatic disease, lines.
BCC might undergo phenotype alterations, subtended — The third group of cell lines was clearly distinct
or not by genetic changes. These alterations would from the two others. It did not express the epithe-
notably include the loss, to a variable extent, of lioid markers found in the ‘luminal epithelial-like’
epithelial-like features, and the gain of more aggres- and ‘weakly luminal epithelial-like’ groups, but in
sive and invasive mesenchymal-like traits. If steroid contrast exhibited a high level of vimentin (gene
receptor-positive cells may occasionally lose their re- VIM), a marker also found in mesenchymal cells.
ceptors (along with other epithelial markers) during Most of these lines had a fibroblastoid phenotype
tumour progression, at least a fraction of ER+ tu- on plastic and grew as colonies with large stellate
mours could evolve to produce ER− metastases. As projections in Matrigel. They were highly inva-
many cell lines have originated from metastatic cells, sive in vitro. BCC lines in this ‘mesenchymal-like’Comparison of cell lines and tumours 257
or ‘stromal-like’ group included MDA-MB-435S, more sporadic than systematic. In fact, the possibility
MDA-MB-231, Hs578T, and BT-549. of an EMT, even transient, has been demonstrated in
two different cell systems: PMC42 and MCF10A. A
Based on this classification, and since: (a) a hall- PMC42 sub-line (PMC42-LA) displays an epithelial
mark of invasive (and metastatic) tumour cells is the phenotype: the cells congregate into pavement epi-
ability to invade and traverse basement membranes; thelial sheets in which E-cadherin and beta-catenin
(b) strong migratory and invasive abilities are also are localised at cell–cell borders. They abundantly ex-
characteristic of cells of mesenchymal origin, it was press cytokeratins, although 5–10% of the cells also
later postulated that BCC with epithelioid features express vimentin. Stimulation of PMC42-LA cells
might acquire a mesenchymal-like phenotype during with epidermal growth factor (EGF) leads to EMT-
tumour progression. This process would be reminis- like changes, including up-regulation of vimentin and
cent of the so-called ‘epithelial–mesenchymal tran- down-regulation of E-cadherin. Vimentin expression
sition’ that occurs during embryonic development at is seen in virtually all cells, and this increase is ab-
precise times and locations [114]. EMT in BCC would rogated by treatment of cells with an EGF receptor
consist of the turning-off of genes encoding epithelial antagonist. Although E-cadherin staining at cell–cell
markers (estrogen receptor-alpha, E-cadherin, tight junctions disappeared in response to EGF, beta-catenin
junction proteins,. . .) and the increase of markers such persisted at the cell periphery. Further analysis re-
as vimentin, accompanied by morphological changes vealed that N-cadherin was present at the cell–cell
and increased invasiveness. In short, cells from the junctions of untreated cells and that expression was
luminal epithelial-like group could evolve to resemble increased after EGF treatment. N- and E-cadherin
cells from the mesenchymal-like group, maybe by ex- are not usually co-expressed in human carcinoma cell
pressing a transitory, weakly luminal, epithelial-like lines but can be co-expressed in embryonic tissues,
phenotype. and this may signify an epithelial cell population
There are data from in vitro studies that support prone to epithelio-mesenchymal-like responses [64].
the EMT hypothesis. For instance, in some MCF-7 MCF10A are not cancer cells, but immortalised nor-
and ZR-75 BCC selected for their resistance to doxo- mal breast epithelial cells. It is, however, of interest
rubicin, vimentin expression appeared to be turned on to mention that transient expression of vimentin may
[115]. Moreover, the MCF-7/Adr cell line, obtained be induced in these cells. In an in vitro wound-healing
by exposure of the luminal epithelial-like MCF-7 to model, analysis of the trajectories of the cells and their
doxorubicin [116], was shown during the course of migratory speeds by time-lapse video microscopy re-
years to express a number of features mainly or ex- vealed that vimentin mRNA and protein expression
clusively found in mesenchymal-like lines such as were exclusively induced in cells at the wound’s edge,
MDA-MB-231 or Hs578T. Experimental expression which were actively migrating towards the center of
of vimentin in MCF-7 BCC led to increased motil- the lesion. Moreover, the vimentin protein disappeared
ity and invasiveness, suggesting that it was needed to when the cells became stationary after wound closure
allow successful invasion [117]. A sub-population, T- [119].
47Dco, was derived from the luminal epithelial-like T- The data from BCC classification [112, 113] are
47D cells. It had unstable vimentin expression and its in agreement with an EMT hypothesis according to
most invasive cells were of fibroblastic/mesenchymal which the weakly luminal epithelial-like phenotype
(VIM-positive) type [113]. Thus, phenotype instabil- could constitute a transitory step in tumour cell pro-
ity or change, resulting in the acquisition of mesen- gression from the luminal epithelial-like portrait to
chymal features, has been observed in BCC lines, and the mesenchymal-like one. However, almost all of the
seems to confer to these cells increased mobility and weakly luminal epithelial-like cell lines in their study
aggressiveness. It has, however, rarely been described have later been found to exhibit specific gene amp-
in vitro. Only a few ER+ cell lines have converted to lifications underlying the over-expression of specific
an ER− phenotype and most efforts to obtain ER− protein tyrosine kinases involved in growth factor sig-
sub-lines from ER+ MCF-7 and T-47D by selection nalling. BT-474, MDA-MB-361, MDA-MB-453, and
or transformation have failed. Even in the cases where SK-BR-3 cells over-express Her-2/neu and have an
a hormone-independence was obtained, the ER was re- amplified ERBB2 locus. This is also observed in about
tained [72, 118]. EMT and more generally, important 30% of tumours. Her-2/neu over-expression has been
phenotype changes in cultured BCC, are likely to be associated to down-regulation of ER and breakdown of258 M Lacroix and G Leclercq
Table 2. A list of genes differentially expressed in BCC lines (at least four cell lines examined) and tumours
Gene name Gene product name(s) Higher expression in References
Luminal ER+ Low-grade
epithelial-like tumours tumours
and/or ER+
BCC lines
Panel Aa
ARHB Ras homolog gene family, member B Yes [122]
C1orf34 Chromosome 1 open reading frame 34 Yes [123]
(DEME-6)
CBX5 Chromobox homolog 5 (HP1 alpha homolog) Yes [124]
CDH1 Cadherin type 1, epithelial cadherin Yes Yes Yes [112, 113, 125]
(E-cadherin)
CLDN1 Claudin 1 Yes [126]
CLDN7 Claudin 7 Yes Yes [127]
DSP Desmoplakin (DPI, DPII) Yes Yes [112, 113, 128]
ESR1 Estrogen receptor, alpha Yes Yes Yes [129–131]
GATA3 GATA sequence binding protein 3 Yes Yes [132]
GPC3 Glypican 3 Yes [133]
GRB14 Growth factor receptor-bound protein 14 Yes [134]
GREB1 Greb1 protein Yes Yes [135]
IGFBP2 Insulin-like growth factor binding protein 2 Yes [136]
IGFBP5 Insulin-like growth factor binding protein 5 Yes [137]
JUP Junction plakoglobin Yes [113]
KLF4 Kruppel-like factor 4 (GKLF) Yes [138]
KRT18 Keratin 18 Yes [139]
MDM2 Mdm2, p53 binding protein Yes Yes Yes [140, 141]
NME1 Protein expressed in non-metastatic Yes Yes [142, 143]
cells (nm23A)
PDZK1 PDZ domain containing 1 Yes Yes [135]
PGR Progesterone receptor Yes Yes Yes [129, 130, 144, 145]
PRDM2 PR domain containing 2, RIZ (transcript 1) Yes [146]
PRLR Prolactin receptor Yes Yes [147, 148]
PTPN6 Protein tyrosine phosphatase, Yes [149]
non-receptor type 6
RERG Ras-like, estrogen-regulated, Yes Yes [150]
growth-inhibitor
SLC9A3R1 Solute carrier family 9, isoform Yes Yes [151]
3 regulatory factor 1
SPINT1 Serine protease inhibitor, Yes [152]
Kunitz type 1 (HAI-1)
ST14 Suppression of tumorigenicity 14 Yes [152]
(matriptase, epithin)
STC2 Stanniocalcin 2 Yes [153]
SYK Spleen tyrosine kinase Yes [154]
TFAP2C Transcription factor activator protein Yes Yes [123]
2 gamma
TFF1 Trefoil factor 1 (pS2, BCEI) Yes Yes Yes [130, 145]
TFF3 Trefoil factor 3 Yes Yes [155]
TJP1 Tight junction protein 1 (ZO-1) Yes Yes [112, 113, 156]
TPD52 Tumor protein D52 Yes [157]Comparison of cell lines and tumours 259
Table 2. (continued)
Gene name Gene product name(s) Higher expression in References
Mesenchymal-like ER− High-grade
ER-BCC lines tumours tumours
Panel Bb
AKT3 V-akt murine thymoma viral oncogene homolog 3 Yes Yes [158]
ANGPT1 Angiopoietin-1 Yes [159]
BZRP Benzodiazepine receptor (peripheral) Yes [160]
CDH3 Cadherin 3, placental cadherin (P-cadherin) Yes Yes [161]
CDH11 Cadherin 11, osteoblast cadherin (OB-cadherin) Yes [162]
CDKN2A Cyclin-dependent kinase inhibitor 1A (p21, Cip1) Yes Yes Yes [163]
CSF1 Colony stimulating factor 1 (M-CSF) Yes [164]
DFNA5 Deafness, autosomal dominant 5 (ICERE-1) Yes Yes [165]
EGFR Epidermal growth factor receptor Yes Yes Yes [55, 144, 166, 167]
ERBB2 c-erb-B2, Her-2/neu Yes Yes [168–173]
ETS1 V-ets erythroblastosis virus E26 oncogene homolog 1 Yes [174]
GPX1 Glutathione peroxidase 1 Yes Yes [175, 176]
GSTP1 Glutathione S-transferase pi Yes Yes [177]
HMGIY High-mobility group protein isoforms I and Y Yes [178]
HXB Hexabrachion (tenascin-C) Yes Yes Yes [179]
IGFBP1 Insulin-like growth factor binding protein 1 Yes [136]
IL6 Interleukin-6 Yes [180]
IL8 Interleukin-8 Yes [181]
IL11 Interleukin-11 Yes [180]
LOX Lysyl oxidase Yes [182]
LOXL2 Lysyl oxidase-like 2 Yes [182]
MET Met proto-oncogene (HGF receptor) Yes [183]
MMP14 Matrix metalloproteinase 14 (membrane-inserted) Yes [184, 185]
MSN Moesin Yes Yes [186]
MT1E Metallothionein 1E Yes Yes [187, 188]
NR3C1 Glucocorticoid receptor Yes [144]
NRG1 Neuregulin 1 (heregulin) Yes [55, 189]
PLAU Plasminogen activator, urokinase Yes Yes Yes [164, 190–192]
PTN Pleiotrophin (heparin binding growth factor 8) Yes [193]
RARB Retinoic acid receptor, beta Yes Yes Yes [194]
S100A4 S100 calcium binding protein A4 (metastasin) Yes Yes [195, 196]
SERPINE1 Plasminogen activator inhibitor type 1 (nexin) Yes Yes Yes [145, 190, 192]
SNAI1 Snail homolog 1 Yes Yes [197]
SNAI2 Snail homolog 2 (slug) Yes [198]
STMN1 Stathmin 1 (oncoprotein 18) Yes Yes Yes [199]
TIMP1 Tissue inhibitor of metalloproteinase 1 Yes Yes [200–202]
VIM Vimentin Yes Yes [112, 113, 179]
a Panel A: genes expressed at higher levels in luminal epithelial-like and/or ER+ BCC; in ER+ tumours; in low-grade tumours.
b Panel B: genes expressed at higher levels in mesenchymal-like/ER− BCC; in ER− tumours; in high-grade tumours.
cell–cell junctions [120, 121], and could indeed par- epithelial-like cell line not studied by Sommers et al.,
ticipate in the phenotype attenuation seen in weakly BT-20 [55], but has been observed in less than 2% of
luminal, epithelial-like cells. For their part, MDA- breast tumours. Finally, MDA-MB-134 cells are char-
MB-468 cells over-express EGFR and have an ampli- acterised by the over-expression of the FGFR1 due to
fied EGFR. This is also seen in another weakly luminal FGFR1 amplification [59]. This event affects 5–10%260 M Lacroix and G Leclercq
of tumours. Among the weakly luminal epithelial-like encoding genes: GATA3 (GATA binding protein 3,
cell lines reported by Sommers et al., CAMA-1 were or GATA-3), KLF4 (Kruppel-like factor 4), TFAP2C
scarcely investigated afterwards, and are the sole lines (AP-2, gamma isoform).
for which the EGFR, ERBB2, and FGFR1 amplifica- Among the genes preferentially expressed in
tion status is unknown. Considering the importance of mesenchymal-like/ER− lines were: SERPINE1 (en-
growth factor signalling in cancer cell properties, it is coding plasminogen activator inhibitor-1), PLAU
speculated that abnormally increased tyrosine kinase (urokinase-type plasminogen activator), and MMP14
activity of Her-2/neu, EGFR, or FGFR1 in BCC could (membrane-type metalloproteinase-1), all implied
significantly alter their phenotype and behaviour. In in proteolysis; ANGPT1 (angiopoietin-1), IL8 (in-
clear contrast to weakly luminal epithelial-like cells, terleukin 8), and MET (hepatocyte growth factor
none of the luminal epithelial-like or mesenchymal- receptor), all related to angiogenesis; the inflamma-
like cell lines described by Sommers et al. has been tion-related genes IL6 (interleukin-6) and IL11
found to exhibit an ERBB2, EGFR, or FGFR1 am- (interleukin-11); genes associated to collagen pro-
plification. As it is unlikely that BCC could progress cessing, such as LOX (lysyl oxidase) and LOXL2
by undergoing an amplification of one of these genes, (LOX-like 2); IGFBP1, HXB (hexabrachion, tenascin-
followed by a return to a normal gene dosage, this may C). All these genes are also frequently expressed
appear to be an argument against the EMT hypothesis. in various mesenchymal cells such as fibroblasts or
However, we cannot exclude the possibility that can- osteoblasts. The mesenchymal-like/ER− BCC lines
cer cells might evolve from the luminal epithelial-like were also characterised by higher amounts of sev-
to the mesenchymal-like phenotype by supporting a eral transcription factors, such as those encoded by
transitory increase in tyrosine kinase activity not sub- HMGIY, ETS1, and SNAI2.
tended by gene amplification. On the other hand, an Thus, from expression analysis of an increasing
EMT could not necessarily include a transition by the number of genes, it appeared that the two phenotypes
weakly luminal, epithelial-like phenotype. that had been previously summarily described [112,
113], the luminal epithelial- and the mesenchymal-
Extended marker analysis. In the course of several like, were indeed highly different. This means that
years, the expression of many genes has been eval- an eventual EMT would imply the turning-off of an
uated in BCC lines. Most of these studies, however, extended set of genes, accompanied or followed by
involved no more than two or three cell lines. When the gain in expression of another wide gene set. Re-
at least four cell lines were examined, they often search has identified a few genes that could play
comprised the two luminal epithelial-like MCF-7 and a key role in regulating numerous other phenotype-
T-47D, and the two mesenchymal-like MDA-MB-231 associated genes. They include CEBPB (encoding the
and Hs578T lines. Such investigations revealed that transcription factor ‘CCAAT/enhancer binding pro-
the expression of the majority of studied genes was tein (C/EBP)beta’), HMGA1 (architectural transcrip-
clearly associated to either one or the other phenotype, tion factors ‘high mobility group protein isoforms I
and positively or negatively correlated to ER expres- and Y’ – HMGI(Y)), ID1 (‘inhibitor of DNA bind-
sion. A series of these genes is mentioned in Table 2(A ing 1, dominant negative helix-loop-helix protein’),
and B). In all cases where it was also measured, the MTA3 (‘metastasis associated 3’, a sub-unit of the
protein amount reflected well the mRNA level. Mi-2/NuRD repressor complex), or SNAI1 (‘snail’, a
Among other genes, luminal epithelial-like/ER+ transcription factor) [203–207]. We cannot exclude the
cells preferentially expressed the epithelium-tied, possibility that the expression of these genes could
serine protease ST14 (matriptase) and SPINT1 (mat- occasionally be more or less deeply altered in BCC,
riptase inhibitor), PRLR (prolactin receptor), SYK either spontaneously or in response to changes in cell
(spleen tyrosine kinase), IGFBP2 (insulin-like growth environment, possibly leading to an at least partial
factor-binding protein 2), IGFBP5, KRT18 (keratin EMT. This remains, however, to be clearly established.
18) and, unsurprisingly, genes up-regulated by acti-
vated ER, such as PGR (progesterone receptor), TFF1 Analysis of breast tumours – markers
(trefoil factor 1/pS2), TFF3 (trefoil factor 3), TPD52 and grade – comparison with cell lines
(tumour protein D52), RERG (Ras-like, estrogen- A number of data have shown that the same ma-
regulated, growth inhibitor). Moreover, these cells jor phenotypic markers distinguishing BCC lines –
had a higher expression of several transcription factor- and clearly associated with the ER status – may alsoComparison of cell lines and tumours 261
discriminate between tumours. Moreover, their ex- bine histological assessment of nuclear pleomorphism,
pression patterns largely overlap histological grade mitotic activity, and tubule formation [210]. Accord-
classification. ing to such systems, tumours classified as ‘grade
I’ or ‘low-grade’ have well-differentiated attributes,
Molecular markers. Part of the genes associated to while ‘grade III’ or ‘high-grade’ tumours have poorly-
either the luminal epithelial-like or the mesenchymal- differentiated attributes. Grade II tumours fall into an
like phenotypes in BCC have also been examined in intermediate category. High-grade DCIS have been
breast tumours. For all of them, their expression was associated with the highest rate of local recurrence
found in at least a fraction of carcinomas. These genes (25–30%), low-grade tumours have very low recur-
are mentioned in Table 2(A) (genes positively cor- rence (0–5%), while intermediate-grade tumours have
related to ER in tumours) and B (genes negatively a recurrence rate somewhere between (10–15%) in a
correlated to ER in tumours). In most cases, the median of 12 years follow-up [211]. Moreover, high-
genes correlated to the luminal epithelial-like/ER+ grade tumours recur within a shorter time than the
phenotype in BCC were positively correlated to the low-grade ones (for instance the median times are 88,
ER expression in tumours, while genes correlated to 42, and 23 months in grades I, II and III, respectively,
the mesenchymal-like/ER− phenotype in BCC were in [212]).
negatively correlated to the ER in tumours. None of As grading is not directly based on molecular ex-
the genes positively correlated to ER in BCC appeared pression profiles, it may be asked whether grades are
negatively correlated to the receptor in tumours. None associated or not to the expression of specific sets of
of the genes negatively correlated to ER in BCC was tumour markers, and more precisely if they are corre-
positively correlated to the receptor in tumours. Thus, lated to the distinct tumour cell phenotypes described
discriminative phenotypic traits observed in BCC lines above for BCC lines and largely retrieved in tumours.
were also frequently discriminative features in tu- It has been repeatedly reported that most ER+ tu-
mours. Regarding ER, it is of interest to note that ESR1 mours are of low-grade. Inversely, high-grade tumours
mRNA variants containing precise truncations in vari- are mainly ER− (see for instance references [129–
ous exons have been identified in tumours; the same 131]). Unsurprisingly, several markers whose expres-
specific variants were also found in BCC lines [208]. sion is positively correlated to that of ER in BCC lines
Although infrequent, the co-existence in the same (and, frequently, in tumours) have also been associated
tumour of markers related to both luminal epithelial- to low-grade. They are mentioned in Table 2, panel
like/ER+ and mesenchymal-like/ER− phenotypes has A. This is the case for CDH1, DSP, MDM2, NME1,
been observed. For instance, ER and EGFR levels are PGR, TFF1, and TJP1. Inversely, high-grade tumours
inversely correlated in BCC and in most tumours, as are characterised by the expression of markers more
shown by numerous studies. Both receptors are, how- related to the ER− profiles in BCC lines and/or in tu-
ever, occasionally co-expressed in carcinomas, but mours. They are mentioned in Table 2, panel B, and
are then, in the vast majority of cases, localised in include CDKN2A, SERPINE1, PLAU, HXB, EGFR,
distinct tumour cells, or in interspersed groups of CDH3, STMN1, RARB (transcript 2), ERBB2. None
cells (‘mosaic expression’, see for instance [209]). of the genes positively correlated to ER in BCC lines
Whether ER-poor/EGFR-rich cells were derived from and/or in tumours was found to be associated with
ER-rich/EGFR-poor BCC, for instance through EMT, high-grade/poorly differentiated carcinomas. None of
in these tumours is unknown. If this was the case, the the genes negatively correlated to ER in BCC lines
observations suggest that these events occur sporadi- and/or in tumours was found to be associated with
cally among cancer cells and do not seem to be related low-grade/well-differentiated carcinomas.
to any significant advantage for progression. Rare co- High-grade DCIS cells also highly expressed sev-
expressions have also been observed with other pairs eral genes encoding extra-cellular matrix proteins and
of markers related to distinct BCC phenotypes (not various growth factors, which likely contributes to the
discussed here). collagen- and fibroblast-rich stroma surrounding these
lesions.
Grade. One of the most widely accepted classifica- The case of ERBB2 is of particular interest.
tion systems for breast carcinomas is grading. The ma- ERBB2-over-expressing (ERBB2+) BCC lines (BT-
jority of grading systems, such as those based on the 474, MDA-MB-361, MDA-MB-453, SK-BR-3)
Scarff, Bloom, and Richardson (SBR) method, com- express markers that make them closer to the262 M Lacroix and G Leclercq
well-differentiated luminal epithelial-like phenotype high between in situ and invasive components of the
than to the mesenchymal-like one (see above). On same tumour. All markers were found to correlate with
that basis, one would have expected to find most grade rather than with invasiveness. No marker was
of the ERBB2+ tumours in the low- or at least the clearly associated with the progression from in situ
intermediate-grade categories. However, while a frac- to invasiveness. The expression of tumour markers
tion of ERBB2+ are indeed of intermediate-grade, was almost identical in the two components of mixed
most of them are of high-grade. For instance, in three lesions [213]. The DNA content and the expression
studies involving hundreds of cases, ERBB2 ampli- of Her-2/neu were simultaneously examined in non-
fication was found in 0%, 10%, and 33%; 3,9%, invasive and invasive phases of primary breast cancers,
20,4%, and 38,9%; 1%, 18%, and 28% of grades I, by image analysis. DNA content in the intra-ductal and
II and III, respectively [170, 171, 173]. Her-2/neu ex- invasive components was virtually identical. Expres-
pression generally reflects ERBB2 amplification. Dis- sion of Her-2/neu was similar in both growth phases,
crepancies between ERBB2 over-expressing tumours implying identity of the Her-2/neu genotype [214].
and cell lines are further discussed in the micro-array In a study of 102 patients, a 67% concordance in
section. grade was found between in situ and infiltrating com-
ponents [215]. Another study of 64 cases indicated
Macroscopic homogeneity of breast tumours – an 86% grade concordance between both components
Stable ‘portrait’ during progression [216]. These studies, and others [217, 218], indicated
a strong correlation between the grade of type of DCIS
According to a common view, progression from and the grade of infiltrating carcinoma in which both
primary to metastatic tumour is accompanied by the components were present.
sequential acquisition of phenotype changes, thus al- It is thus striking that patterns of grade or the other
lowing BCC to invade, disseminate, and colonise markers did not seem to change during the transition
distant sites. Based on in vitro data, it has notably been from in situ to invasive carcinoma. Invasive cancer
proposed that BCC in vivo might undergo a transition seems to occur independently of tumour grade. This
from the luminal epithelial-like to the mesenchymal- is further supported by comparative genetic hybridisa-
like phenotype. Along the same lines, it has been tion data (see below).
repeatedly suggested that tumour progression is char-
acterised by a shift from well differentiated/low-grade Recurrence, metastasis. Metastatic and recurrent
to poorly-differentiated/high-grade category. Never- BCC appear late in tumour progression. They are
theless, most investigations have revealed that pro- believed to have accumulated alterations since their
gression is not accompanied by major changes in initial transformation event. On the other hand, meta-
marker expression or grade. static cells may colonise various tissues often highly
different from the breast after having completed all
Progression to invasiveness and markers/grade. If steps of a complex process including local invasion,
changes in phenotype/grade were frequent during intravasation, resistance to blood pressure, adhesion
progression from in situ to invasive carcinoma, it to blood vessels and extravasation. This suggests that
should then be easy to find in a significant part they have sequentially acquired specific adaptive prop-
of invasive tumours both luminal epithelial-like and erties. All this supports the hypothesis that metastatic
mesenchymal-like markers, and both low- and high- and recurrent cells could have a phenotype signif-
grade compartments. Logically, mesenchymal-like icantly different from that observed in the primary
markers and high-grade areas should be more often tumour. This supposition is of high importance,
observed in the invasive than in the in situ tumour since many BCC lines originated from metastatic
compartment. In fact, most studies examining this cells.
point have revealed a striking similarity between both Attempts have been made to compare the expres-
parts of breast carcinomas [138, 213–218]. For in- sion of various markers and/or histological grade in
stance, histopathological grading and tumour marker primary tumours and their corresponding metastases
(p53, Her-2/neu, Ki-67, ER, PgR, bcl-2 and angiogen- and/or recurrences. It was shown that KRT8 and
esis) expression were compared in 194 pure DCIS, 127 KRT19 expression was similar in both primary car-
small invasive lesions, and 305 lesions with both an in- cinomas and their lymph node (LN) metastases [219].
vasive and in situ component. Grade concordance was In an immuno-histochemical (IHC) study of 38 LNYou can also read
